Publication

Abstract : Various systems were developed to assist those with vision impairments and enhance their standard of life. Regrettably, the majority of these systems have restricted functionality. We decided to build a model that detects all possible surrounding things and guides them through voice description. Our main objective is to detect the obstacles in the person’s pathway and warn them using voice description [1]. The model can assist blind people from the surrounding obstacles which exist in their pathway. The primary goal of the model is to identify the nearby obstacles - vehicles, animals, persons, objects, etc, in the person’s pathway and guide or warn them using voice description. Our proposed system integrates advanced real-time obstacle detection and voice-guided navigation using a novel combination of DepthNet module with convolutional LSTM layers, enhancing spatial context awareness compared to traditional models. This approach, along with fine-tuning on diverse datasets such as MS COCO and BDD100K, ensures robust performance in varied environments and conditions. The real-time guidance capability is further improved by the integration of optical flow algorithms, which accurately estimate motion in dynamic scenes. This comprehensive solution significantly enhances the assistive technology for those with visual impairments.